Add fold optimization for `push N dup`

[lionkor]

Code such as

push 1
dup

should be optimized to

push 1
push 1

or similar, depending what is faster. The main benefit would be that constant folding will hit in the latter case, if followed by, say, add.

Consider an example similar to that from the readme as of time of writing:

push 42
dup
mul

Currently, this wouldn't be constant folded to push 84, however this optimization would allow folding like that.